The result is a system that can take requests by voice to determine a blind person's location and offer audible directions to the user's destination of choice. While the current prototype is limited to the University of Florida campus -– using a spatial database containing walkways, buildings, roads, light poles, benches and other objects –- Moore told NewsFactor that he hopes to have a commercial product available for real-world use within the next two years.

Technology Mix

Moore said the DRISHTI system uses hardware consisting of the Xybernaut MAV wearable computer, Trimble DGPS (differential global positioning system), Motorola iDEN phone or Lucent WaveLan Wireless PC card, and Sun Microsystems Ultra 60 workstation.

As for software, the wearable computer uses IBM ViaVoice and custom code written with the Java API (application program interface) and Java Communications API to interface with the GPS receiver. The workstation uses ESRI's ArcView mapping software, more custom software with Java API, and the Oracle 8 database management system.

Database Directions

Using the spatial database of the university's campus, the system "sees" the user's location –- which is obtained via GPS receiver –- and relation to walkways, streets and obstacles, Moore explained. "The user's spatial relationship to the objects in the database is continuously monitored, and the user is notified with synthesized voice prompts when he is approaching an obstacle or is deviating from his path," he said.

While the numerous tall buildings of a typical city would prohibit such GPS-based locating, Moore said another method, such as a camera system, could be integrated with a spatial database to achieve the same result. "Many cities have some sort of GIS (geographic information system) to keep track of their resources already," he noted.

Next Challenges

Moore's blind father provided input for the system and tested it on the UF campus. The student said the next step for the DRISHTI includes a more conversational communication style, rather than the command and response of the prototype.

"As of now, the interaction with the computer is [intended] to get a time-sensitive job done," he said. "You don't want too much conversation, as the recognition accuracy goes down, and the goal is achieved [more slowly] -- but a better help language is being designed."

Assisted on the project by University of Florida computer engineering professor Sumi Helal and civil and coastal engineering doctoral student Balaji Ramachandran, Moore said the biggest challenges to producing a commercial product for the blind are having "a pure vocal user interface-oriented operating system" and maintaining a spatial database that allows users to input and store objects they encounter on their journeys.